Agreed. I think the dialogue wheel -- with its "noble/paragon" response near the top, its "smartass/sarcastic" response in the middle, and its "direct/renegade" response near the bottom -- can be extremely limiting. That's because we're choosing our responses based largely upon the attitude we wish our hero to express instead of on the complex idea we wish to communicate. That's not how conversations work in real life.
Yes, attitude is important, but when someone asks you a question you don't immediately think, "Should I answer this in a pleasant, cheeky or angry manner?" That'd be absurd. No -- most of the time, you listen carefully to the speaker and then you frame a logical response that can vary in several ways based upon the information you've just received. The tone of voice you use to deliver your response is a mere by-product of the idea you've chosen to convey.
That's why I think BioWare should replace the simplistic, misleading, emotion-based paraphrases it currently employs with full sentences that express complex ideas. And the tone of voice in which these complex ideas are delivered should be readily apparent from the way the sentences are written.
In my experience, RPG players like to read, they like think, and they like to have complete control over the characters they create. The dialogue wheel in its current form makes this impossible.
I feel that a deeper conversation system like this would give us a far richer roleplaying experience wherein we can create a character of genuine nuance and substance, as opposed to a character of childish emotional extremes.
I agree to an extent. The preconceptions we as players have about the dialogue wheel influence the conversation's course too much in its (the wheel's) most recent iterations. However, I don't think that's down to a dichotomy between reason and emotion as you argue.
First and foremost, the gameplay incentive for uniform characterization and metagaming conversations in general needs to be removed. In the first Mass Effect some of the reputation checks were absurdly high for no clear narrative or gameplay reason, like the Ethan Jeong persuasion on one of the earliest story worlds, while also detracting from combat readiness. It was tyrannical in Mass Effect 2 without an import or save editing and was also a problem to a lesser extent in Dragon Age II. The worrisome thing is that while Mass Effect 3's reputation system is an improvement on this, the devs also eviscerated the neutral option and stuffed Shepard to the gills with autodialogue and autocharacterization ("Thessia's lost, and that's on me"). Two steps forward, one step back.
Second, the different options need to stack up meaningfully against each other in terms of narrative and characterization. The problem with the neutral option in both games of Mass Effect in which it exists is that it doesn't really do much to characterize Shepard, even as a nonchalant or noncommittal personality. It's basically a way to advance the conversation to the next branch without engaging with the other party's dialogue. In this respect, DA does a better job, because the middle option is both neutral in tone (demurring with wit doesn't hard commit to a position) and colors the protagonist in a certain way (while also allowing NPCs to respond in kind). The ME wheel's neutral option never gained much of an identity not only because the trilogy's morality system doesn't have a role for it, but also because it's just throwaway writing.
Third, the wheel is too much of an abstract object of its own at this point. It presumably started out as cognitive shorthand for communicating the tone, mood and emotion of spoken language through written text. However, now it exercises an independent influence on how we perceive the things the protagonist says and does (is something paragon or renegade because the game says it is?) and limits the ways in which he can be characterized by the writers. In a more general sense, this is the course of all institutions.
Also as a sidenote, I don't agree that "most of the time," human communication comprises "fram[ing] a logical response that can vary in several ways based upon the information you've just received." I truly wish that were the case. However it seems more realistic (and generally accepted in social science) that a lot of human interaction is indeed based on emotion, the confirmation of priors, the fulfillment of preferences, and even baser obligations like tiredness and hunger.